Carlos J. Vilalta y Perdomo wrote: > > Would you please give me some advice in the following problem? > Problem: > Pearson correlation of x1 with y is n.s. > Yet, when y = x1 + x2 + x3... + x6 + e, then x1 becomes a significant > predictor of y > Questions: > 1. Would this be effect of an intervining/mediating variable? > 2. What would you do? Include or exclude x1 from the analysis? > I appreciate your comments and suggestions. > Carlos
x1 is more highly correlated with the residuals of the model with the other variables included than with y on its own. I believe this a type of supression effect. Whether to include x1 depends on many other factors. For example, what is x1, why did you correlate it with and so forth. In a large data set you are bound to find such effects, so simply trawling through different models is not a good strategy. OTOH if there is theory to suggest x1 might influence y it might be sensible to have it in the model. Thom . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
